ummary

Introduction

In 1986 the Health 2000 Report, a long term health policy document, was presented to the Dutch parliament. This document is part of shift in interest in public health towards health rather than he& services planning. There are two interesting features in this shift. The one is the tendency to measure the effectiveness of a policy, an intervention or a technology in terms of health, the outcome rather than the input, output or process. The other is the acceptance that political choices need to be made, since however large the budget for health is, it will always be limited.

One of the choices to make will be whether or not to invest in preventive interventions. Preventive interventions can be defined as deliberate changes in the prevalence of risk factors in a population. To be able to weigh the costs and the benefits of such preventive interventions, an estimate will have to be made of their effect on the health of the population. Furthermore changes in risk factor prevalence may also occur autonomously. An estimate of the changes in the health status of the population as a result of these shifts in risk factor prevalence, will be important for the planning of health services and for the setting of realistic targets, as proposed by WHO.

Prevent is a tool that will estimate the health effects of changes in risk factor prevalence in a population, as a result of trends or interventions. Its results can either be used directly in health policy making to formulate targets or quantify different scenario’s on changes in risk factor prevalence in the future, or its results can be used as input for formal decision making processes such as for instance cost effectiveness studies.

The Prevent model

In epidemiology an analysis of the distribution of disease incidence and risk factor prevalence in different populations is used to confirm the hypothesis

251

252 SUMMARY

of a causal relationship between risk factor and disease. The strength of the relationship is often expressed as the ratio of incidence between exposed and non exposed, the Incidence Density Ratio (IDR). The importance of a risk factor for the incidence of a certain disease in a population is usually expressed as the Etiologic Fraction (EF), the proportion of the total inci- dence of the disease that can be attributed to the prevalence of that risk factor in the population. The EF is sometimes used as an indication of the proportion of incidence that could be prevented by the total elimination of that risk factor in the population.

However, since most often prevention will not eliminate but merely re- duce the prevalence of a risk factor, a measure was developed to estimate the impact of a change in prevalence of a risk factor on the incidence of a disease, the Potential Impact Fraction (PIF). It stands for the incidence that is avoided by a preventive intervention as a proportion of the incidence that would have occurred in that population without the intervention.

Prevent estimates the effect of changes in risk factor prevalence in a population in terms of health benefit. It is based on the epidemiologic effect measure the Potential Impact Fraction. To achieve the objectives of the project it has incorporated the following three requirements in the methodology:

?? the possibility that one risk factor affects several diseases, and that one disease is affected by several risk factors,

?? a time dimension to simulate the reduction in excess risk after cessa- tion of exposure to the risk factor,

?? the interaction between the effect of the intervention and the demo- graphic evolution in the population.

The model will simulate the development, over time, of two populations: one as a result of trends in risk factor prevalence and demography, and the other which incorporates both trends and interventions on risk factors and the demography. Differences between these two populations are the effect of the intervention. In the current version of Prevent all measures of health benefit are based on mortality.

Results

Since the results are primarily intended for policy making the choice of the disease categories to include in the project was determined by criteria, relevant to public health policies:

SUMMARY

?? the disease had to contribute significantly to the ill health of the Dutch population,

?? the disease should have known risk factors upon which interventions could reasonably be applied.

The availability of reliable data both on the relationship between risk factor and disease incidence and on the prevalence of the risk factor in the Dutch population, determined which risk factors and diseases were included in the current version of the model. Sensitivity testing on the input data and a historical reconstruction of lung cancer mortality in the Netherlands were done to provide an indication of the robustness of the model.

With the Prevent model a number of risk factor interventions were sim- ulated and the results analyzed. The interventions show that there is a con- siderable difference in the outcome measured as a proportional mortality reduction as with the Potential Impact Fraction and an outcome measure in absolute terms such as mortality reduction. Especially for a common cause of death such as Ischemic Heart Disease even a small proportional reduction may represent a large number of deaths prevented.

Interventions on the prevalence of smoking in the population illustrate both the importance of a multifactorial approach and of the time dimension. If only one disease is considered much of the effect of the intervention is missed. In order to evaluate the effect of a risk factor intervention the total (aggregated) health b enefit should be considered. This also includes the possible increase in causes of death not related to the risk factor considered, as a result of the intervention.

The timelag between smoking cessation and the ultimate effect on lung cancer mortality illustrates the consequences of the introduction of a time dimension. Not only does a considerable period elapse before the full effect of a reduction in smoking prevalence can be appreciated, but it also means that the quite sizable reductions in smoking prevalence in the recent years will be noticeable in an initial reduction in lung cancer mortality in the near future in both the trend and the intervention population. With the introduction of a time dimension changes in risk factor prevalence in the past will continue to affect health in the future.

In the Dutch population which can expect a large increase in the pro- portion of elderly in the coming years, the absolute number of cases of diseases which occur mostly in old age, will increase sharply in the future. For some diseases even considerable reductions in risk factor prevalence will not be able to counteract this increase. This means that even with a suc- cessful preventive policy the need for certain curative services will continue to increase. If demography is not taken into account reductions in mortal-

254 SUMMARY

ity rates as a result of a preventive intervention may create the erroneous impression that the absolute number of cases will also go down.

When comparing the effect of different risk factor interventions it is important to keep the following items in mind:

?? Not only the total health benefit should be compared but also the distribution over sub groups in the population.

?? The simulation time over which interventions are compared should be sufficiently long to show the full effect of each intervention.

?? The choice of benefit measure will influence the priority setting.

Of all the risk factors in the Prevent model, the greatest health benefits are to be expected from a reduction in smoking prevalence. Not only a new generation of non smokers will greatly improve the populations health, a program of smoking cessation will also result in sizable mortality reductions, although mostly for men.

Finally the Prevent model was applied to a number of recent Dutch policy documents to show how it can provide useful quantitative informa- tion for policy making. The alternative smoking scenario’s of the Dutch Lifestyle Scenario project were simulated with different assumptions of the population groups affected by the intervention. It shows that the health benefits will greatly differ depending on the group in the population in- tervened upon. A similar quantification of the hypothetical smoking inter- vention analyzed in the Dutch Cancer Scenario, illustrates why it is useful to look at risk factor interventions in a multifactorial model since the ex- pected benefits are much higher if other diseases affected are also taken into account.

The alternative risk factor reductions necessary to achieve the targets as stated in the Health 2000 Report and the more recent policy document on the Prevention of Cardiovascular Disease, were calculated. For Ischemic Heart Disease it will be impossible to achieve the target with an interven- tion on one risk factor only. Even in a multifactorial intervention large reductions in risk factor prevalence in the near future, will be necessary if the target is to be achieved before the year 2000.

Prevent shows a positive aspect of the interrelationship of risk factors and diseases. Some interventions proposed to achieve one disease specific target may automatically achieve another disease target. This is the case in all interventions suggested for the reduction in Ischemic Heart Disease: the inherent reduction in smoking prevalence ensures that the lung cancer target is achieved without additional interventions. This illustrates the necessity to apply targets in a comprehensive health policy, and not only by disease category.

SUMMARY 255

Conclusions

The results of hypothetical interventions simulated by Prevent show that a multifactorial model with a time dimension will yield different estimates of the effects of risk factor intervention than the traditional epidemiological measures. The use of a dynamic population model makes it easier to visu- alize the interaction between changes in age specific mortality (as a result of changes in risk factor prevalences) and demography. This is especially important for policy making since it helps to show that despite a possible increase in absolute disease specific mortality, the mortality reduction due to a preventive intervention can be quite large.

Prevent only expresses health benefits in outcome measures based on mortality. In the future it would be useful to extend this to morbidity measures also. This will only be possible if curative care is not assumed static over time. The model should therefore be extended into a public health model which not only looks at preventive interventions but also at interventions on curative care. It will then be possible to also express effects in terms of changes in the utilization of services or even costs.